Date:
Monday, November 23, 2020, 10:30am to 11:30am
Location:
Zoom - please contact emilie_campanelli@fas.harvard.edu for more information
Title:
Global Convergence of EM?
Abstract:
In this talk I will first discuss a recent joint work with Yihong Wu: https://arxiv.org/abs/1908.10935. We show that the randomly initialized EM algorithm for parameter estimation in the symmetric two-component Gaussian mixtures converges to the MLE in at most $\sqrt(n)$ iterations with high probability. Then I will mention the limitations of that work and propose an extension to general Gaussian mixtures by overparameterization.